GO-2
AgiBot GO-2 (Genie Operator-2) is an AI behavioral foundation model and successor to GO-1, serving as the software brain for AgiBot physical humanoid robots. Part of AgiBot broader software stack that includes GE-2 (action world model), Genie Sim 3.0 simulation, and AimRT middleware. Not a physical robot platform itself but the AI system running on AgiBot A-series and G-series hardware. GO-3 with ViLLA architecture is expected in Q3 2026.
Primary use case: AI foundation model for humanoid robot control; generalist physical intelligence for AgiBot robot fleet including manipulation, navigation, and task execution
Height
75 cm
Weight
15 kg
DOF
12
Specifications
Physical
Height
75 cm
Weight
15 kg
Degrees of Freedom
12
Locomotion
bipedal
AI & Compute
AI Capabilities
facial recognition, autonomous navigation, obstacle avoidance, speech interaction, persona customization, route guidance, exhibition explanation
Commercial
Availability
commercial
Primary Use Case
AI foundation model for humanoid robot control; generalist physical intelligence for AgiBot robot fleet including manipulation, navigation, and task execution
Related news
The FCC just cleared its first foreign factory robots, from ATi Robotics
Swiss startup Embodied AI launches AIR, a humanoid robot that learns from factory workers
Agility is public: the short list of humanoid robot stocks, and who's next
Toyota Robotics: Toyota just established a dedicated robotics division
The China humanoid shakeout begins: 150 companies at risk
Breakthrough Google Robots: Master Garbage Bag Tying, Kettle Handling & Light Bulb Replacement – Next-Level Mind-Blowing AI Robotics Innovation
Google DeepMind launched Gemini Robotics 2, featuring three models (ER 2, VLA, and On-Device 2) that enable whole-body intelligence for humanoid robots, allowing them to think and reason while moving and operating simultaneously rather than sequentially. The system was tested on Apptronik Apollo 2 and Franka Duo, demonstrating capabilities like picking up kettles, tying garbage bags, screwing in light bulbs, and multi-robot collaboration, though success rates for complex dexterous tasks remain between 32-92%.
Google Robotics Team Achieves Major Feat: Solving the Long-Standing Industry Challenge of the "Last Few Centimeters"
Google DeepMind unveiled Gemini Robotics 2, a general-purpose robot model featuring whole-body intelligence that coordinates legs, torso, arms, and fingers of humanoid robots to perform complex tasks like walking while bending, picking up objects from the ground, and multi-finger dexterous operations including screwing light bulbs and tying garbage bags.
AgiBot Refreshes Lingxi X2 for Guided Tours, Retail and Entertainment
AgiBot Refreshes Lingxi X2 for Guided Tours, Retail and Entertainment